TOPIC:

Some of my customers in Finland have rejected UMIK-1 as measurement tool for loudspeaker design because it can't be used in 2 ch measurement with electrical reference signal. I have been solving this problem for a while and I think I got the solution figured out.

I have sold ESI U24 XL interface among UMIK-1 as simple sound card that offers analog outputs and both coaxial and optical digital outputs. It has very good connectivity in most cases. The interface has also the same connections in as out and ASIO-support for drivers.

I use Win10 myself and it is possible to connect multiple USB-interfaces with ASIO4ALL on Windows to REW. This makes it possible to connect the electrical loop cable on ESI's analog connection, set the loop output and input on REW to those and direct the main measurement signal to ESI's output while measuring input is UMIK-1.

There might be a problem that the digital delay on system varies when rebooting or restarting REW, but it seems like it is stable during one session which should be enough. I did some measurements today and they are behind the link . Here is a screenshot of impulse responses of different measurements on the same microphone position. The floor noise from my room varies a little. There is also a measurement where the microphone has been moved about 50 cm closer to the speaker. They seem to have the same timing.

Interesting idea, but I don't understand the measurement. The timing reference should remove the delay due to the computer, so that the impulse peak represents acoustic delay time only. In your measurement, the impulse delay is 223 ms.
?

Good point. The measurement (231 ms) includes delay from UMIK-1 and ASIO4ALL input drivers.

Distance between microphone and speaker was about 1,5 meters and 231 ms is about 76 meters (344 m/s). This is the problem, that the internal delay from UMIK-1 and input drivers varies on different sessions. It seems that the measurements are comparative when they are made during one session. I haven't tried yet different sessions, but this is what I have been told.

From ASIO4ALL settings I tapped the hardware buffers to be used. In case loudspeaker design is the use for the measurements , the internal delay among acoustic delay doesn't matter, because you only need to compare the impulse response and its timing between the drivers you are using and all the measurements have the same extra delay. Of course the impulse timing can be adjusted from the measurements and the extra delay can be removed.

This is just the only way to measure timing and phase accurate with UMIK-1 so that the data can be used on precise loudspeaker modeling for example with LspCAD or VituixCAD.

john.reekie wrote:

Interesting idea, but I don't understand the measurement. The timing reference should remove the delay due to the computer, so that the impulse peak represents acoustic delay time only. In your measurement, the impulse delay is 223 ms.
?

Actually the extra delay is UMIK-1 internal delay - (minus) ESI U24 XL internal delay. Let me know, if that's not correct.

Different day and different delays. Though now the internal delays seem more correct.




Any way, this isn't a problem if you measure all you need during one session.

For all the ones interested on this, I can share new measurement data for REW.

aaninen.fi/media/2ch-measuring/


I booted and changed the USB-ports between three measurements. Seems like the inner delay varies depending on which USB-ports has been used. My laptop has three and I changed between two ports.

There is more information on the notes of the measurements.

Thanks much for sharing this method Jussi!

USB on any works asynchronously, so depending on when you're starting your measurement (and the state of the buffer), it will have varying delay. Hence the reason why REW introduced the timing reference since 5.15
www.minidsp.com/applications/auto-eq-wit...measuring-time-delay

DevTeam

The acoustic reference signal isn't usable when you are measuring drivers for loudspeaker design. It can be used only when you adjust two or more loudspeakers together. Its accuracy isn't enough when you need to know the exact phase and timing behavior of different drivers. Electric loop cable is needed and REW supports this kind of measurements.

Problem is that UMIK-1 isn't advised to be used in this situation cause no analog connection is available for second channel input and loop cable. My approach for 2 ch measuring with the equipment shown before is a solution for this situation. A well known loudspeaker designer Kimmo Saunisto from Finland has developed a CAD for loudspeaker design, VituixCAD and he has lack of interest on UMIK-1 cause it hasn't been possible to use it for 2 ch measurements. He has great doubts also on my solution for it, but I myself have been using it successfully.

With UMIK-1 and ESI U24XL the real delays can't be measured but the measured phase and delay between different measurements are accurate. This makes the measurements usable for loudspeaker design. There are some buts though which are aught to be taken in consideration.

Loop back in output device only is able to send timing reference to measurement application from output device only so assumption about repeatability of timing+phase measurement by mic is based on constant latency between reference input and mic input. Those are two different USB devices in proposed setup so driver should be able to synchronize input streams from different USB input devices in all situations. I didn't get evidence about that.

Acoustic reference hanging in the mic boom at constant distance from mic capsule might work okay enough, but it's limited to REW. One risk is that mic boom leads sound wave sometimes mechanically more than acoustically so connection should be lossy. Other locations for acoustic reference do not support mic movements which are usually easier than speaker movements while polar measurements of different drivers in the construction.

The most easiest and stable system is standard measurement microphone (i.e. other than Umik-1) connected to dual channel soundcard with phantom and loop back cable to reference input channel. The same device captures both mic and reference. Amount of components (including cables) is equal so there is really no reason to play with suspicious or actually unreliable system.

Any how I have a little car audio project going on where adjusting the delays on each channel is very important. There is a C-DSP 8x12 v2.0 in my use and the whole setup is fully active. Crossovers for each driver has been done with DSP.

At first I adjusted the delays by my ears, but it didn't get me satisfied. I got UMIK-1 in the right position quite well and used ESI U24XL with it as soundcard to provide the measurement signals for the C-DSP 8x12 and to the other channel of ESI's input as reference signal. ASIO4ALL drivers are the only way to get this done with Room EQ Wizard.

The problem of variety in the internal delays on different USB devices is real in case you need to boot up you computer or change the host ports of your devices. In my opinion these cases need to be noticed, but the 2 ch measuring works during one session when you don't change the host ports of devices. Of course the measured delays aren't real values cause of the internal delays of USB devices, but different measurements during one session are comparable to each other. This makes it possible to measure all channels in you car system and adjust the delays exactly.

Here is overlays of impulse responses including left tweeter and left mid bass in my car, if I remember correctly. Delays are adjusted already to match up. The delay adjustment of five micro seconds on the processor was a little compromise according to the measurements. So it seems like this way of 2 ch measuring is accurate enough.




What I have been thinking lately is that drivers which produce different frequency ranges show up different kind of impulse responses like in the overlays I posted. For car audio measuring on channels delays my idea is to limit the measuring range for about an octave around the crossover frequency to be used. This would make it easier to read the impulse response diagrams on the overlays. I'll get back to this idea later.